gemma3-4b-new-exp

This model is a fine-tuned version of google/gemma-3-4b-it on the new_exp_train dataset. It achieves the following results on the evaluation set:

• Loss: 0.1973

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 0.0002
• train_batch_size: 1
• eval_batch_size: 1
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 4
• optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: cosine
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 5

Training results

Training Loss	Epoch	Step	Validation Loss
0.4136	0.6452	20	0.3526
0.1675	1.2903	40	0.2498
0.1294	1.9355	60	0.2083
0.0626	2.5806	80	0.2021
0.0338	3.2258	100	0.1973
0.0244	3.8710	120	0.2242
0.0109	4.5161	140	0.2246

Framework versions

• PEFT 0.15.2
• Transformers 4.52.4
• Pytorch 2.10.0+cu128
• Datasets 3.6.0
• Tokenizers 0.21.1